Difficulty in Convergence of Fluid in Conjugate Heat Transfer

[Krystalfang]

Dear experts，

I'm recently trying to make a model of a cold plate with heat sources, radiant elements, power, cold plate, S-shaped channel. This is about 33w grids. I want to achieve fluid-heat coupling via conjugate heat transfer.

However, in the calculation process, I found that nonlinear iteration is difficult to converge, but linear iteration is easy to converge. Basically, each step of nonlinear iteration can only be calculated in two steps before jumping out of the loop, and the decline rate can reach 10-8. And the nonlinear The main problem of iteration focuses on the fluid. When I do a rough mesh, the fluid doesn't have that much mesh and it converges very quickly.

# Fluid properties
mu = 0.001   # Pa*s
rho = 998.2    #kg/m3
cp = 4128      #j/kg-k
k = 0.6     #w/m-k


[Mesh]
  type = FileMesh #Read in mesh from file
  file = liure2.e
[]


[Variables]
   [T_s]
    block = '1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18'
    initial_condition = 293
  []
  [T1]
    block = '2'
    initial_condition = 293
  []
   [./velocity]
    family = LAGRANGE_VEC
    block='2'
  [../]
  [./p]
    order = FIRST
    family = LAGRANGE
    block='2'
  [../]
[]

[ICs]
  [velocity]
    type = VectorConstantIC
    x_value = 1e-15
    y_value = 1
    z_value = 1e-15
    variable = velocity
  []
[]


[Kernels]
  # Diffusion kernel for each block's variable
  [diff_0]
    type = ADHeatConduction
    variable = T_s
    block ='1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18'
  []
  [fluid_conduction]
    type = ADHeatConduction
    variable = T1
    block = '2'
    thermal_conductivity = 'k'
  []
  [source]
    type = HeatSource
    variable = T_s
    value = 1e8
    block ='3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18'
  []

 [./mass]
    type = INSADMass
    variable = p
  [../]
  [./mass_pspg]
    type = INSADMassPSPG
    variable = p
  [../]
  [./momentum_convection]
    type = INSADMomentumAdvection
    variable = velocity
  [../]
  [./momentum_viscous]
    type = INSADMomentumViscous
    variable = velocity
  [../]
  [./momentum_pressure]
    type = INSADMomentumPressure
    variable = velocity
    pressure = p
    integrate_p_by_parts = true
  [../]
  [./momentum_supg]
    type = INSADMomentumSUPG
    variable = velocity
    velocity = velocity
  [../]
 [./temperature_advection]
   type = INSADEnergyAdvection
   variable = T1
 [../]
  [temperature_supg]
    type = INSADEnergySUPG
    variable = T1
    velocity = velocity
  []
[]

[InterfaceKernels]
   [fluid_to_plate_10]
    type = ConjugateHeatTransfer
    variable = T1
    T_fluid = T1
    neighbor_var = 'T_s'
    boundary = '2'
    htc = 5000
  []
[]

[AuxVariables]
  [T]
  []
[]

[AuxKernels]
  [temp_0]
    type = NormalizationAux
    variable = T
    source_variable = T_s
    block = '1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18'
  []
  [temp_1]
    type = NormalizationAux
    variable = T
    source_variable = T1
    block = '2'
  []
[]

[BCs]
 [./plane_wall]#PCB
  type = ADConvectiveHeatFluxBC
  variable = T_s
  boundary = '1'
  T_infinity = 293
  heat_transfer_coefficient = 7.62
  [../]
  [./vec_inlet]
    type = VectorFunctionDirichletBC
    variable = velocity
    boundary = '3'
    function_y = 1  #m/s
  [../]
  [./inlet_temp]
    type = DirichletBC
    variable =T1
    boundary = '3'
    value = 293
  [../]
  [./outlet_p]
    type = DirichletBC
    variable = p
    boundary = '4'
    value = 0
  [../]
[]

[Materials]
  [./const_fluid]
    type = ADGenericConstantMaterial
    prop_names = 'rho mu cp k'
    prop_values = '${rho} ${mu}  ${cp}  ${k}'
    block = '2'
  [../]
  [ins_mat]
    type = INSADStabilized3Eqn
    velocity = velocity
    pressure = p
    temperature = T1
    block='2'
  []

  [./plate]
    type = ADHeatConductionMaterial
    thermal_conductivity = 202.4    # W/m k
    specific_heat = 871  # J/kg k
    block = '1'
  [../]

  [./fluid]
    type = ADHeatConductionMaterial
    thermal_conductivity = .6  # W/m k
    specific_heat = 4128    # J/kg K
    block = '2'
  [../]


  [./source]
    type = ADHeatConductionMaterial
    thermal_conductivity = 54    # W/m k
    specific_heat = 330    # J/kg k
    block = '3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18'
  [../]
[]

[Preconditioning]
  [Newton_SMP]
    type = SMP
    full = true
    petsc_options_iname = '-pc_type -pc_factor_mat_solver_package'
    petsc_options_value = ' lu       mumps'
    solve_type = 'NEWTON'
  []
[]

[Executioner]
  type = Steady
  l_max_its = 5
  nl_max_its =1000
  nl_rel_tol = 1e-8
  l_tol = 1e-5
  petsc_options_iname = '-pc_type -pc_factor_mat_solver_package -ksp_gmres_restart'
  petsc_options_value = 'lu       superlu_dist                  200'
  automatic_scaling = true
  scheme = bdf2
[]
[Debug]
  show_var_residual_norms = true
[]

[Outputs]
  exodus = true
  file_base = liure_out3
  perf_graph = true
[]

Also, what does the htc stand for in conjugate heat transfer, since it seems to me that conjugate heat transfer doesn't require the definition of the convective heat transfer coefficient, I'm confused

[GiudGiud]

Hello

I would try reducing the flow rate to make it an easier problem and see if you can get convergence
You can also turn remove part of the physics, like the conjugate heat transfer, and see if you get better convergence that way

linear convergence working fine is not surprising given that you are using LU, a direct method.

Guillaume

[Krystalfang]

Hi Guillaume
Thank you very much for your reply, I think your statement is correct. I have also tried this recently, and I found that when I set the flow rate to 0.1m/s, MOOSE can converge, but when I try to adjust the flow rate to 0.5m/s, it does not converge and appears the following question:

This problem is also encountered in general simulation software, but how to solve this problem in MOOSE,because the flow velocity required by this model is 2.6m/s. I would like to hear your suggestions.
In addition, as to what you said to remove the conjugate coupling module, I will try further.
Krystal

[GiudGiud]

For a zero pivot error, you should add this parameter to your preconditioning
-pc_factor_shift
NONZERO

You can also experiment with different LU solver packages